A Nonlinear Feature based Epileptic Seizure Detection using Least Square Support Vector Machine Classifier

Epilepsy is the most common disease of central nervous system. According to World Health Organization about 50 million people worldwide and 80% people from developing regions are suffering from epilepsy. Electroencephalogram (EEG) is one of the non-invasive techniques available for seizure detection. In this paper we have proposed non-linear feature based epileptic seizure detection using least square support vector machine (LSSVM) classifier. We have developed low computational and more accurate system for real time epileptic seizure detection. Symbolic entropy, Lempel-Ziv complexity and sample entropy are extracted and LSSVM classifier is used to classify data into ictal, healthy and inter-ictal EEG signals. LSSVM classifier in One-verse-All approach, One-verse-One approach, and multi-class classifier approach classifies ictal EEG signal with an accuracy of 81.67%, 91.25 % and 82.22 % respectively. Hence, the proposed One-verse-One approach has detected ictal EEG signal with highest accuracy and sensitivity.